What happens to biological data inside GEPER.

One variant's journey

1. 01

   Biological data

   What comes in?

   Genomic variants, together with the context needed to make sense of them. A variant is treated as an object with a location, a molecular neighbourhood and a population history, not an isolated string.

   • Sequences and variants
   • Gene, transcript and protein context
   • Population frequency and conservation
   • Curated biological and clinical knowledge
2. 02

   Computational representation

   How does biology become computable?

   Inputs are normalised and enriched into consistent, structured descriptions, so that two scientists or two models asking about the same variant start from the same definition of it.

   • Standardised variant description
   • Structured biological annotation
   • Sequence- and protein-level representations
   • Missing information stays explicit
3. 03

   Modelling

   What do the models do?

   Machine learning and statistical models estimate the potential effect of a variant. Each model is an instrument with known strengths and blind spots; no single output is treated as a conclusion.

   • Sequence-based machine learning
   • Statistical and population-genetics methods
   • Established computational-biology techniques
   • Outputs are evidence, not verdicts
4. 04

   Interpretation

   How does evidence become insight?

   Model estimates are combined with curated knowledge under published interpretation frameworks. The reasoning is recorded so that a result can be audited: what contributed, how strongly, and what could not be assessed.

   • Published interpretation frameworks
   • Recorded reasoning and provenance
   • Explicit “not assessed” states
   • Built for expert review
5. 05

   Output

   What comes out?

   A readable interpretation intended for qualified reviewers: the conclusion, the evidence behind it, and its limits, presented so that it can be checked and challenged.

   • Conclusion with supporting evidence
   • Uncertainty stated plainly
   • Designed for human sign-off